332\_lab-jev-chat
Android overlay that reads on-screen chats, uses Jev to judge intent, risk and whether to reply now, then ranks drafted replies.
Overview
Jev Chat Assistant is a non-invasive conversation helper that sits on top of chat apps. It reads the conversation currently on screen through Android accessibility (WeChat and mobile QQ work; Lark capture is partly done), without hooking the app or using its API. It never sends messages or touches payments; it only fills the input box.
Jev does the judging. One request asks seven questions covering the other person's real intent, a danger level, their needs, the best action and whether to reply right away, returning in about a second. A generative model (DeepSeek by default, via OpenRouter) drafts three candidate replies, and Jev ranks them. Results show in a translucent overlay where you copy or fill a reply and press send yourself. Adapters per app make it portable.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What is the other person's real intent? | — |
| Score | How dangerous is this conversation? | 1-9 scale |
| Yes / no | Should I reply right now? | yes · no |
| Choice | Which of the three drafted replies is best? | — |
How it works
- Accessibility adapter extracts title and messages from the foreground chat
- Jev answers seven judgment questions in one request
- A generative model drafts three replies
- Jev ranks the drafts
- Overlay shows results; user copies or fills a reply and sends manually
Strengths
- No app API or account access needed
- About one-second judgments per the README
- Human always presses send
Caveats
- Disguises its accessibility service name to read WeChat nodes, which may break with app updates
- Lark message body capture still in progress
- Needs OpenRouter and Jev keys
At a glance
- Type
- App
- Status
- prototype
- Language
- Kotlin
- Runtime
- Android app
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
The Windows app reads visible WeChat chat text via UI Automation or local OCR, uses Jev for structured intent judgment, and optionally uses DeepSeek to generate three copyable reply suggestions.
logicrw/awesome-jev-projects
Frequently asked questions
Does it send messages for me?
No, it only fills the input box; you press send.
What does Jev decide?
Intent, danger level, needs, best action, whether to reply now, and the ranking of drafted replies.
Which apps work?
WeChat and mobile QQ on Android end to end; Lark partly.
Related projects
Android WeChat assistant showing Jev emotion probabilities and intent readings beside messages; never sends replies.
Appbeta- jev-chat-windows★ 802
Windows WeChat reply assistant: local offline OCR reads the chat, Jev judges intent and tension and ranks three drafted replies to fill in.
ProjectproductionPython AstrBot plugin where Jev decides whether the bot should join a group conversation; the main model still writes the reply.
ProjectproductionPythonDeepSeek Harness plugin that classifies each user message's emotion, urgency and department with Jev and injects it as context.
PluginbetaTypeScriptiOS custom keyboard that shows a copied message's intent, risk level and candidate replies inside any chat app, filling but never sending.
AppbetaSwiftWindows desktop chat assistant that OCRs a selected chat region, has Jev analyse the conversation, and ranks drafted replies for copying.
ProjectbetaPython, TypeScript